Multiscale polymorphic uncertainty quantification based on physics-augmented neural networks

doi:10.1016/j.cma.2025.118726

Multiscale polymorphic uncertainty quantification based on physics-augmented neural networks

2026-01-10

0

OA

AI

DOI：10.1016/j.cma.2025.118726

Original

Save

View PDF

Abstract

En 中文

• consideration of polymorphic uncertainty is enabled for homogenization on all scales. • Uncertainty from meso- to macroscale by convergence of representative volume element. • Numerically feasible uncertainty analyses require physics-augmented neural networks. • Uncertain input quantities are considerable in physics-augmented neural networks. • Interval probability-based random fields characterize spatial variation of structure.

Keywords:

Homogenization

Uncertainty quantification

Multiscale analysis

Multiscale uncertainty quantification

Polymorphic uncertainty

Physics-augmented neural networks

Multiscale polymorphic uncertainty quantification based on physics-augmented neural networks

Abstract

AI Summary

Journal

Organization